Critique 129: Association of alcohol consumption with skin cancer; A report from the Women’s Health Initiative — 20 November 2013
Kubo JT,
Authors’ Abstract
PURPOSE: The relationship between alcohol consumption and preference of alcohol type with hazard of melanoma (MM) and risk of non-melanoma skin cancer (NMSC) was examined in the Women’s Health Initiative (WHI) Observational Study (OS).
METHODS: A prospective cohort of 59,575 White postmenopausal women in the WHI OS (mean age 63.6) was analyzed. Cox proportional hazards models and logistic regression techniques were used to assess the hazard and risk of physician-adjudicated MM and self-reported NMSC, respectively, after adjusting for potential confounders including measures of sun exposure and skin type.
RESULTS: Over 10.2 mean years of follow-up, 532 MM cases and 9,593 NMSC cases occurred. A significant relationship between amount of alcohol consumed and both MM and NMSC was observed, with those who consume 7+ drinks per week having a higher hazard of MM (HR 1.64 (1.09, 2.49), p global = 0.0013) and higher risk of NMSC (OR 1.23 (1.11, 1.36), p global < 0.0001) compared to non-drinkers. Lifetime alcohol consumption was also positively associated with hazard of MM (p = 0.0011) and risk of NMSC (p < 0.0001). Further, compared to non-drinkers, a preference for either white wine or liquor was associated with an increased hazard of MM (HR 1.52 (1.02, 2.27) for white wine; HR 1.65 (1.07, 2.55) for liquor) and risk of NMSC (OR 1.16 (1.05, 1.28) for white wine; OR 1.26 (1.13, 1.41) for liquor).
CONCLUSIONS: Higher current alcohol consumption, higher lifetime alcohol consumption, and a preference for white wine or liquor were associated with increased hazard of MM and risk of NMSC.
Forum Comments
Background: Data on the relation between the risk of skin cancers associated with alcohol intake and smoking are mixed A case-control study by Westerdahl et al of 400 Swedish subjects in 1996 revealed no influence of alcohol or smoking on the risk of melanoma (Westerdahl J, Olsson H, Måsbäck A, Ingvar C, Jonsson N. Risk of malignant melanoma in relation to drug intake, alcohol, smoking and hormonal factors. Br J Cancer 1996;73:1126–1131). However, Freedman et al (2003) in a report from the US Radiologic and Technologists study of more than 65,000 subjects found that alcohol consumption of > 14 drinks/week increased the risk of melanoma (RR 2.1), while smoking decreased the risk (RR 0.6) (Freedman DM, Sigurdson A, Doody MM, Rao RS, Linet MS. Risk of melanoma in relation to smoking, alcohol intake, and other factors in a large occupational cohort. Cancer Causes Control 2003;14:847-857).
Two studies in 2004 (Naldi et al, Millen et al) had opposite results, with the former showing no effect of alcohol on skin cancer risk while the latter suggesting an increase in risk with high alcohol consumption (Naldi L, Gallus S, Tavani A, Imberti GL, La Vecchia C; Oncology Study Group of the Italian Group for Epidemiologic Research in Dermatology. Risk of melanoma and vitamin A, coffee and alcohol: a case-control study from Italy. Eur J Cancer Prev 2004;13:503-508) (Millen AE, Tucker MA, Hartge P, Halpern A, Elder DE, Guerry D 4th, Holly EA, Sagebiel RW, Potischman N. Diet and melanoma in a case-control study. Cancer Epidemiol Biomarkers Prev 2004;13:1042-1051). In 2012, Song et al reported a decrease in melanoma cases with smoking, as found in many previous studies (Song F, Qureshi AA, Gao X, Li T, Han J. Smoking and risk of skin cancer: a prospective analysis and a meta-analysis. Int J Epidemiol 2012;41:1694-1705).
In a recent (2013) presentation to the European CanCer Congress (ECCO) in
Comments on the present study: In the present study, the amount of alcohol consumed was reported for 4 periods: ages 14-17, 18-22, 23-29, and 30-49 years; these values were then used to estimate the lifetime intake in drink-years. The authors considered the frequency of the average number of drinks/week in categories of non-drinker, past drinker, <1 drink/month, < 1 drink/week, 1-7 drinks/week, and 7+ drinks per week.
At baseline, more frequent drinking was noted among leaner subjects (BMI <25) and less so for obese subjects. Subjects who were active had higher intake than sedentary subjects. Subjects reporting the highest sun exposure, those who stated “burns and tans,” and college-educated subjects reported higher alcohol intake. As has been shown in many previous studies, the risk of skin cancers were lower among smokers than among non-smokers.
This was a fairly large study, based on 532 cases of melanoma (MM) and 9,593 cases of non-melanoma skin cancer occurring over approximately 10 years of follow up among more than 59,000 women in the Women’s Health Initiative. The key reported results were for a higher hazard of MM (HR 1.64) and NMSC (OR 1.23) for drinkers of 7 or more drinks/week, compared with non-drinkers. Lifetime alcohol consumption was also positively associated with hazard of MM and risk of NMSC, with the only significant increases for MM according to type of beverage seen for women with a preference for white wine or liquor. The authors also report that the risk of these cancers was lower among smokers than among non-smokers.
It is noted that there were large decreases in the estimates of HR associated with alcohol intake when adjusted for potential confounders. For example, the HR for MM decreased for total wine from 1.30 to 1.06 when adjusted; for red wine:1.71 to 1.34; for white wine: 1.93 to 1.52; for liquor: 1.87 to 1.65; for beer: 1.34 to 1.18. The magnitude of these changes, as well as the very different baseline characteristics of the women according to their drinking habits, raises a strong possibility that there may be residual confounding affecting the results. In particular, there may have been large differences in SES among subjects (which relates to risk of skin cancer), and this may not have been adjusted for adequately in the analyses.
In his review of this paper, Forum member Zhang stated: “Overall the analysis methods are appropriate and I can’t see anything wrong. I have two minor comments:
“(1) If the study is to assess the relation of alcohol consumption to the risk of melanoma, I was wondering why the main analysis included subjects with history of melanoma (n=844 vs. n=532 occurring during follow-up) and NMSC (prevalent n=5,705 vs. n=9,593 occurring during follow up). I noticed that in the sensitivity analysis the investigators did exclude the individuals with a history of NMSC and MM, and the analysis excluding those with a history of MN or NMSC showed similar results for baseline consumption and baseline preference, with elevated HRs/ORs for the highest category of consumption compared to non-drinkers, as well as for white wine and liquor drinkers compared to non-drinkers.
“(2) In the main analysis, adjusting for history of MN or NMSC in the regression model may lead to potential collider bias, if we believe that alcohol consumption, especially lifetime alcohol consumption, may be associated with (or cause) MN or NMSC. However, such bias, in general, might tend to dilute the effect of alcohol consumption on recurrent or new MN or NMSC.
“(3) I agree that the very different baseline characteristics of the women according to their drinking habits raises a strong possibility that there may be residual confounding affecting the results.”
Forum member Klatsky had further comments on his own recent analysis and the implications of the present paper. “In the Kaiser-Permanente study, among persons preferring wine the HR at 3+ drinks per day was 1.7 (1.2-2.5), while it was 1.2, 1.3, and 1.1 in persons with preference for liquor, beer, and no beverage type. Among heavy drinkers, only the wine relation was significant, but the CI’s overlapped. At 1-2 drinks per day only those with no preference had an increased risk (HR = 1.4) and the wine drinkers had a HR of 0.9. So maybe we should not make too much of the beverage type data in these analyses. However, the relation with alcohol seems statistically solid.
“I do think that there are enough reports to indicate an association of alcohol consumption with melanoma. The present paper has a nice discussion of what little is known about hypothetical mechanisms and confounding. But I am puzzled by fact that they do not cite the report by Mukamal (Mukamal KJ. Alcohol consumption and self-reported sunburn: a cross-sectional, population-based survey. J Am Acad Dermatol 2006;55:584-589). With data from 300,000 subjects participating in a risk factor surveillance survey, Mukamal found that “Approximately 33.5% of respondents reported a sunburn within the past year. Heavier average alcohol use and binge drinking were both positively associated with the prevalence and number of sunburns within the past year. The adjusted odds ratios for prevalence and number of sunburns among binge drinkers were 1.39 (95% confidence interval 1.31-1.48) and 1.29 (95% confidence interval, 1.20-1.38), respectively.” Klastky wonders whether “possible confounding by sun exposure is not only an artifact of SES and tropical vacations, but whether heavy imbibers at the beach may fall asleep when exposed or simply do not notice their sunburn until it is severe.” Forum members Van Velden, de Gaetano, and Svilaas agree that “Ultraviolet irradiation is still the major causative factor of these cancer, and remains as a confounding variable in most studies.”
Reviewer Klatsky adds: “However, even if the alcohol-melanoma association is due to confounding by sun exposure, the same explanation seems less plausible for the lower risk of melanoma associated with smoking. While there is obviously no public health utility in any ‘protective’ effects of smoking on skin cancer, that association is scientifically interesting and could be a clue to mechanisms.”
Reviewer Finkel had an interesting comment: “The heaviest drinker tend to occupy settings not illuminated by the sun, the major inciter.” It is not known how many really heavy drinkers were included in the highest category of drinking (all reporting 7+ drinks/week were in the same category). While this group reported high sun exposure, it is assumed that this group may well contain many subjects consuming only slightly above 7 drinks/week as well as some very heavy drinkers. With such a mixture of drinkers, adjusting for sun exposure may be problematic.
Forum Summary
A report from more than 59,000 women in the Women’s Health Initiative related reported alcohol consumption to the risk of melanoma (MM) and non-melanoma skin cancer (NMSC). This was a fairly large study, with 532 cases of melanoma and 9,593 cases of NMSC occurring over approximately 10 years of follow up. The key reported results were for a higher hazard of MM (HR 1.64) and NMSC (OR 1.23) for drinkers of 7 or more drinks/week, compared with non-drinkers. Lifetime alcohol consumption was also positively associated with hazard of MM and risk of NMSC, with the only significant increases for MM according to type of beverage being seen for women with a preference for white wine or liquor. As shown in many previous studies, the risk of these skin cancers was lower among smokers than among non-smokers.
Forum reviewers thought that this was a very well-done analysis. They did note that there were large decreases in the estimates of HRs related to alcohol consumption when adjustments were made for sun exposure and other known confounders. For example, the HR for MM decreased for total wine from 1.30 to 1.06 when adjusted; for red wine:1.71 to 1.34; for white wine: 1.93 to 1.52; for liquor: 1.87 to 1.65; for beer: 1.34 to 1.18. The magnitude of these changes, as well as the very different baseline characteristics of the women according to their drinking habits, raises the possibility that there may be residual confounding affecting the results. In addition, reviewer Zhang points out that by including subjects with these skin cancers prior to baseline in their main analyses, there could be a problems with bias in their results.
Forum member Klatsky has recently reported on this topic using several decades of follow-up data from the Kaiser-Permanente study, with more than 300,000 subjects and more than 1,000 cases of MM. In that study, he and his colleagues found that among persons preferring wine, the HR for MM at 3+ drinks per day was 1.7 (95% CI 1.2-2.5), while it was 1.2, 1.3, and 1.1 in persons with preference for liquor, beer, and no beverage type. However, at 1-2 drinks per day, wine drinkers had a HR of 0.9. Klatsky states “So maybe too much should not be made of the beverage type data in these analyses. However, the association with alcohol seems statistically solid.”
A previous study by Mukamal, with data from 300,000 subjects participating in a risk factor surveillance survey reports: “Approximately 33.5% of respondents reported a sunburn within the past year. Heavier average alcohol use and binge drinking were both positively associated with prevalence and number of sunburns within the past year. The adjusted odds ratios for prevalence and number of sunburns among binge drinkers were 1.39 (1.31-1.48) and 1.29 (1.20-1.38), respectively.” Klatsky wonders whether “possible confounding by sun is not only an artifact of SES and tropical vacations, but whether heavy imbibers at the beach may fall asleep when exposed or simply do not notice their sunburn until it is severe.” Other Forum reviewers also worry that there may be residual confounding in the present study from sun and ultraviolet exposure. Klatsky adds: “However, even if the alcohol-melanoma association is due to confounding by sun exposure, the same explanation seems less plausible for the inverse association between smoking and melanoma. While there is obviously no public health utility in any ‘protective’ effects of smoking, that association is scientifically interesting and could be a clue to mechanisms.”
The bottom line is that there are considerable observational epidemiologic data suggesting that alcohol consumption may relate to an increase in the risk of MM and NMSCs. As mechanisms are not known, there is still concern that much of this relation may relate to residual confounding by ultraviolet sun exposure, the most important environmental factor for these diseases.
* * *
Comments on this critique were provided by the following members of the International Scientific Forum on Alcohol Research:
Arthur Klatsky, MD, Dept. of Cardiology, Kaiser Permanente Medical Center, Oakland, CA, USA
David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa
Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA
Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway
Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA
Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy
R. Curtis Ellison, MD, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA